X-ArecaNet: Dataset of arecanut X-ray images for deep learning applications

The grading of arecanut using a non-destructive approach specifically through internal examination remains an unexplored area of research. We present an X-ray image-based dataset of Arecanuts for grading or classification. Density is one of the key metrics used to determine the quality of Arecanut when grading is performed manually. Low-quality Arecanuts are less dense, allowing X-rays to pass through easily, resulting in black patches on the X-ray image. In contrast, denser Arecanuts block X-rays, appearing as white patches in the X-ray image. The objective of this paper is to present a novel dataset of Arecanut X-ray Images for grading in Arecanut industry. and establish a benchmark for quality inspection standards. Our dataset consists of X-ray images of 3 grade types: Grade1, Grade2, and Grade3 in the decreasing order of quality. The images in dataset are distributed into 'training' and 'validation' folders, with80 % allocated for training and 20 % allocated for validation.They are annotated in YOLOv5 format. A total of 900 images are collected representing 300 images for each grade type. The image augmentation is not performed for the images in dataset. We believe that the provided dataset is beneficial for designing, developing, evaluating, and validating a deep learning models for Arecanut grading and are applicable for setting a benchmark for quality check of an Arecanut in the agriculture industry.